1. Using language technology for SNOMED CT localization
Stefan Schulz, Jose A Miñarro-Gimémez
Examples for redundancy of sub-terms (number of descriptions in which phrases occur) :
Background
Goal
"Magnetic resonance imaging"
627
"off-road motor vehicle, except snow vehicle"
351
"second degree burn"
166
The absence of localized clinical terms mapped to SNOMED CT concepts constitutes a major barrier for SNOMED CT adoption. The translation of Fully Specified Names alone is debatable, due to the mismatch with the terms clinicians commonly use.
The alternative is a bottom-up, incremental creation of interface terms connected to SNOMED CT IDs. Although this is not a translation in the first place it could also be used as raw input to speed up principled SNOMED CT translation efforts.
A tentative workflow for semi-automated acquisition of interface terms linked to SNOMED CT is presented and described for the German language. It shows the potential of semi-automated term generation methods capitalizing on three factors:
redundancy of SNOMED CT content
word similarities
Web-based machine translation.
Examples for term similarities acreoss languages (here English - German) :
"cryoglobulinemia"
"Kryoglobulinämie"
"cardioesophageal"
"kardioösophageal"
"chemothalamectomy"
"Chemothalamektomie"
Web-based machine translation
Current Status
The process is currently being refined and tested. First results suggest its usefulness for speeding up the localization of SNOMED CT content. Another possible output would be the creation of a raw translation in the process of professional SNOMED CT translation which could be taken as input by professional translators who build an official SNOMED CT translation. The main use case for the German raw interface terminology is text mining.
The application of this process for translating an interface terminology from Spanish to Portuguese is planned.
Acknowledgements: European Commission SEMCARE – 7th FP grant ; ASSESS-CT - H2020-PHC , grant
Poster presented at SNOMED CT Expo 2015, Montevideo, Uruguay, October 29 – 10, © Institute for Medical Informatics, Statistics and Documentation, Medical University of Graz, Auenbruggerplatz 2, 8036 Graz, Austria –